218 



EXPERIMENTAL DESIGN 



Fig. 16-1. A completely random- 

 ized design. V, L, S stand for sep- 

 arate treatments; the locations were 

 chosen by chance. 



the three groups is to be placed here ( 1 or 4 for soil, 2 or 5 for vermiculite, 

 3 or 6 for liquid, for example). With a second throw, decide which plant 

 of the four is to be placed here. Proceed to the second available spot and 



repeat the process. Continue until all 

 twelve plants are assigned. If by 

 chance all the plants in soil should be 

 together, trust the die because dice are 

 better judges of randomness than you 

 are. The plants in Fig. 16-1 have been 

 labeled S, V, and L for soil, vermicu- 

 lite, and liquid, respectively. They 

 were placed by exactly the method de- 

 scribed. There are simpler ways of 

 randomizing, but this serves as an ex- 

 ample. 



Now we allow the plants to grow 

 until they have produced large leaves 

 and then cut off the leaves and weigh them. The statistical analysis tells 

 us whether there are any significant differences among the three treat- 

 ments. 



Randomized Block: Use the same example as above. We may suspect 

 that one end of the bench is very slightly warmer than the other end 



or that light intensity and humidity are 

 not quite uniform. In order to distrib- 

 ute the environmental variation more 

 evenly among the plants, we divide 

 the area into four "blocks" of three 

 spaces each. One plant from each treat- 

 ment goes into each block, but within 

 the block the plants are randomized. 

 One such finished design is shown in 

 Fig. 16-2. 



Latin Square: This is a modified 

 randomized block which helps to over- 

 come environmental variation in two 

 directions. Suppose that temperature 

 and light intensity vary in the plant 

 growth room, not only from end to 

 end, but from side to side as well. In this instance we arrange the space 

 in rows and columns, making sure that each treatment occurs in each 

 row and in each column. The Latin square need not be square, but, for 



V. V3 L3 S, 



S3 L, V3 L, 



1-4 S, S, V, 



Fig. 16-2. A randomized block 

 design. Each vertical block includes 

 each treatment once. Within the 

 block the treatments are random. 

 The blocks could also be arranged 

 horizontally. 



